Forty-one shea tree populations were sampled, spanning the main climatic zones of Vitellaria paradoxa Gaertn. in Mali and 10-35 adult trees were chosen randomly in the agroforestry parklands of each area. A total of 12 morphological traits, related to tree morphology, fruit size and leaf form were measured. The variance components showed that variation among populations represented the smaller percentage of the total variation with most of the values varying between 15 and 30%. The repeatability coefficient was generally high for tree within populations with values ranging between 0.23 and 0.78. Although genetic correlations cannot be accurately estimated, due to difficulties in separation from environmental effects, the results indicate that there is a very low genetic relation between the three kinds of traits, i.e., between those related to tree, those related to leaf and those related to fruit. Leaf and fruit size traits were positively and significantly correlated with rainfall, although tree circumference was negatively correlated with rainfall and the significantly larger shea trees were noted in the drier areas -an observation thought linked to human management of the parklands. Soil drainage and parkland density, however, did not explain differences between populations for fruit traits. This study offers preliminary information for the development of a breeding population for a shea tree improvement programme. The value of repeatability, the low correlation between sets of traits and the distribution of variation, suggest that selection of many individual trees within a few populations, would allow capture of large genetic gain especially for fruit traits.

The relationship between trees, grass and soil in a dry savanna in Mali was investigated, to identify variables that are most relevant to assess vegetation units. A 65 ha plateau was inventoried using a systematic square grid sampling pattern. Thirteen soil or topography variables, and tree and grass characteristics were measured at each sampling point. Multivariate analysis was used to separately analyse soil, tree and grass data, and to characterize tree-grass and tree-soil relationships. Four units of soils, four units of tree formations, and four units of grass formations were identified. There was a correspondence between these groups, indicative of four vegetation units: thicket, bare land, shrub savanna and tree savanna. Soil depth and soil texture were the soil variables that best related to tree vegetation. A negative correlation was found between tree basal area and grass dry biomass. Finally, vegetation units, as identified from tree species composition, had contrasted diameter structures and densities. RésuméLa relation entre les arbres, les herbes et la terre dans une savane sèche du Mali fut enquêtée, afin d'identifier des variables qui sont plus pertinentes dans l' évaluation des unités végétales. Un plateau de 65a fut recensé utilisant un modèle d'échantillon systématique par quadrillage. Treize variables de terre ou topographie, ainsi que les caractéristiques des arbres et des herbes furent mesurées à chaque point échantillon. Des analyses multivariées furent employées afin d'analyser individuellement les données venant de la terre, des arbres et des herbes, et afin de caractériserles rapports arbre-herbe et arbre-sol. Quatre unités de terre, quatre unités de la formation des arbres, et quatre unités de la formation de l'herbe furent identifiées. Il existait une harmonie entre ces groupes, indicative de quatre unités de végétation: broussailles, terre dénudée, savane arbrisseau, et savane forestière. La profondeur et la texture de la terre furent les deux variables qui agissaient le plus surla végétation arboricole. Une corrélation négative fut trouvée entre la taille de l'étendue basale de l'arbre et la biomasse d'herbes sèches. Finalement, les unités végétales, identifiées par la composition d'espèces d'arbres, déployaient des structures de diamètre et de densité contrastantes.

No abstract

No abstract

No abstract